The hydrolysis of geminal ethers: a kinetic appraisal of orthoesters and ketals

A novel approach to protecting jet fuel against the effects of water contamination is predicated upon the coupling of the rapid hydrolysis reactions of lipophilic cyclic geminal ethers, with the concomitant production of a hydrophilic acyclic hydroxyester with de-icing properties (Fuel Dehydrating Icing Inhibitors - FDII). To this end, a kinetic appraisal of the hydrolysis reactions of representative geminal ethers was undertaken using a convenient surrogate for the fuel–water interface (D2O/CD3CN 1:4). We present here a library of acyclic and five/six-membered cyclic geminal ethers arranged according to their hydroxonium catalytic coefficients for hydrolysis, providing for the first time a framework for the development of FDII. A combination of 1H NMR, labelling and computational studies was used to assess the effects that may govern the observed relative rates of hydrolyses.


Introduction
Our interest in organic dehydrating agents for applications in the aeronautical industry is stimulated by the commercial case for developing a new approach to managing water contamination in jet fuel. The acid-catalysed hydrolysis reactions of cyclic orthoesters present an attractive platform from which to develop Fuel Dehydrating Ice Inhibitors (FDII) [1], since jet fuel is itself mildly acidic [2] and the products of hydrolysis can in principle afford protection against ice formation by residual water [3]. From Brønsted's ground-breaking work on acid catalysis [4], to more recent investigations as models for glycosidic bond cleavage [5], orthoesters have been examined using a range of methods, solvents (i.e., water, methanol, dioxane, and mixtures thereof), and conditions [6,7]. It has been difficult to draw upon this disparate body of data for our purposes, as potential FDII will operate under non-buffered, relatively apolar conditions. Though easily stated, the mechanism for the formation of a charged intermediate [5,[7][8][9][10][11] followed by attack of water, cleavage of RO-C bonds, and several proton-transfer reactions Values of k H+ determined via calibration with respect to 8 [18] and 16 [23], respectively (see Experimental section).

Scheme 1:
The three-stage mechanism for the specific acid-catalysed hydrolysis of cyclic orthoester A.
The first stage sees the generation of 1,3-dioxolan-2-ylium cation B along with alcohol; as water is usually in excess the reverse reaction with ROH (i.e., k −1 ) is considered negligible. It has been established that cyclic orthoesters, and in particular those derived from 1,3-dioxolanes, initially undergo exclusive cleavage of the exocyclic alkoxy group [19]. In reviewing the case of orthoesters, Kresge et al. suggests two mechanistic extremes [19,24]. A concerted catalysed process, wherein the -OR group undergoes protonation as the C···O(H + )R bond begins to undergo cleavage. The alternative stepwise process involves the C-OR bond spontaneously cleaving at a rate similar to that of RO − protonation by H 3 O + . The latter describes a spontaneous uncatalysed hydrolysis. In the case of 1,3-dioxolanes, the concerted mechanism is believed to dominate [25]; however Guthrie [9] asserts that orthoesters are in fact delicately poised between stepwise and concerted processes; depending upon the substituent, both mechanisms are operational for aryl dimethyl orthoformates [21]. Stage 2 sees reaction of B with H 2 O to afford 2-hydroxy-1,3-dioxolane C (i.e., k 2 ) with the overall equilibrium constant Finally, stage 3 sees the acid-catalysed (i.e., k 3 [H + ]) cleavage of C to afford hydroxy ester D.
For acyclic geminal ethers stage 1 is invariably rate limiting, i.e., k 3 > k 1 . For cyclic systems k −3 becomes more dominant in the pH range of about 4-6 [23], however stage 1 remains rate limiting [26]. The overall rate of reaction can therefore be established by measuring the consumption of the geminal ether [27]. We present here kinetic data measured for a range of acyclic orthoformates, orthoacetates, 1,3-dioxolane orthoesters, oxanes, and 1,3-dioxanes (Figure 1), and consider the factors which may modulate the rates of hydrolyses.

Results and Discussion
The ratio of rates corresponding to the hydrolysis of some sixmembered ketals and orthoesters were reported previously [18]. Employing a similar approach using the same solvent system (i.e., D 2 O/CD 3 CN 1:4 v/v), quantitative rate data for the hydrolysis reactions of acyclic and cyclic derivatives 1-16 were de-termined, affording a library of potential FDII, organised unambiguously by hydroxonium catalytic coefficient k H+ (Figure 1, k H+ M −1 s −1 in parentheses; see Supporting Information File 1 for further details). The observed rate constants (k obs ) were determined using the rate equation integrated over time, using 1 H NMR spectroscopy to evaluate relative concentrations at 25 ± 0.5 °C. Having already established [1] that specific acid catalysis is more likely to be the dominant mechanism operating in jet fuel/water mixtures, the complications attending the use of buffered systems (and potential k 3

Acyclic orthoesters
Previous workers compared the rates of hydrolysis for cyclic 1, 4, 8 and 15 as determined in D 2 O/CD 3 CN (1:4 v/v) with acyclic derivatives whose rates were established in water [18]. Further, the literature suggests that acyclic 3 and 6 hydrolyse at similar rates (i.e., k H+ = 1.4 and 1.2 × 10 4 M −1 s −1 , respectively) [4,7] which is surprising given the charge stabilisation expected to accompany the replacement of an H atom with a methyl group. A comparison of the relative rates of pairs of acyclic orthoesters (i.e., 2 + 3, 2 + 6, and 3 + 13) was performed to address such inconsistencies. An eleven-fold increase in the hydroxonium catalytic coefficient accompanies the replacement of the MeOwith EtO-moieties in orthoformates 2 → 3 ( Figure 1). A threefold increase in k H+ accompanies MeO → EtO for orthoacetates 6 → 13. As expected, alkyl substituents on the carbonyl carbon atom have a greater accelerating influence on the overall rate of hydrolysis (H → Me ca. 50-fold) compared to the ethereal substituent (Me → Et 3-11-fold), with the fastest rate of hydrolysis for acyclic systems considered here being achieved by 13 (k H+ = 20.1 ± 0.9 M −1 s −1 ).

The relative reactivity of six-membered cyclic ketals and orthoesters
The ratio of rates for hydrolysis (as opposed to k H+ ) for 1, 4, 8 and 15 have been reported previously [18]. We prepared 8 and evaluated k H+ using the same conditions employed by these workers to calibrate values of k H+ for 1, 4, and 15 ( Figure 1). Acyclic 6 was also examined alongside 8 to correlate the hydroxonium catalytic coefficients of the hydrolysis reactions of cyclic and acyclic derivatives in this study. It is clear that exo-cyclic orthoester 15 is the faster reacting geminal ether of the series, hydrolysing at five-times the rate of constitutionally isomeric 8 ( Figure 1). The kinetic anomeric effect was invoked previously to rationalise the relative reactivity of 15 with respect to endo-cyclic 8 [18]. One would anticipate a significant change in rate to accompany the reduction of charge-stabilising oxygen atoms within 8 to afford 4; yet k H+ (8)/k H+ (4) ≈ 13:4. The antiperiplanar lone pair hypothesis (ALPH) proposes that the axial anomer of 4 constitutes the major conformer in solution [28], perhaps affording some stereoelectronic advantage to an early transition state which appears operative in the case of such acid-catalysed processes [29]. The stereoelectronic advantage [30] of an endo-cyclic oxygen atom in 4 may mitigate the loss of an electrostatically stabilising oxygen atom from the system. The hydroxonium catalytic coefficient for the hydrolysis of 1 is ≈650 times slower than observed for the constitutional isomer 4. Acid-catalysed cleavage of the former affords a leaving group covalently tethered to a cation which renders the overall rate apparently slow, perhaps through a favoured re-cyclisation. It is noteworthy that the hydroxonium catalytic coefficient for the hydrolysis -albeit measured in water -of a similar yet acyclic ketal (i.e., 2,2-diethoxypropane) is several orders of magnitude greater than 1 [18].

Relative reactivity of cyclic orthoesters; five versus six-membered rings
As k 1 ≈ k obs in the pH range examined here (Scheme 1), factors associated with the relief of cyclic strain cannot be used to account for the difference observed for five-membered 5 and six-membered 8 (i.e., k H+ = 6.5 and 9.8 M −1 s −1 ). Further, the hydroxonium catalytic coefficient for the hydrolysis of 5 and acyclic analogue 6 are within experimental error of each other (i.e., k H+ = 6.5 ± 0.2 and 7.0 ± 0.2 M −1 s −1 , respectively). It was noted previously that the relative rates of hydrolysis for sixmembered 15 and 8 could be explained with the kinetic anomeric effect. Consistent with this, the X-ray crystal structure of an analogous yet conformationally constrained bicyclic orthoester possesses an unusually elongated axial C─O bond (Figure 2a), which undergoes preferential cleavage with Lewis acids [31]. The Cambridge Structural Database (2015) [32] contains a single example of a five-membered 1,3-dioxolane orthoester [33]. Here, the ring adopts a distorted half-chair (C 2 ) arrangement with a dihedral angle θ  (5)]. This, along with a rate of hydrolysis similar to an acyclic system suggests that a kinetic anomeric effect does not extend to 1,3-dioxolane orthoesters.
The reactivity of C(4/5) substituted 1,3-dioxolanes Substitution of the 1,3-dioxolane ring at C(4/5) introduces asymmetry with the attendant challenges of isomer separation  We attribute the two-fold rate increase of 10-11 (k H+ ≈ 12 M −1 s −1 ) with respect to 5 (k H+ = 6.5 M −1 s −1 ) to be steric in origin. The additional two-fold rate acceleration of 14 (k H+ = 27.9 M −1 s −1 ) with respect to 10-11 is ascribed to the relief of transannular compression involving the C(4/5)-Me and Me-C(2) groups, which nOe studies suggest are near to each other in space (see structure 14, Figure 3).

The reactivity of C(2) substituted 1,3-dioxolanes
McClelland et al. [23] has determined k H+ for 5, 7, 9, 12 and 16 in a water/phosphate buffer system at pH 6-7; we prepared 5 and 16 to calibrate their data with the conditions employed here (i.e., D 2 O/CD 3 CN/HCl), whilst assuming that the relative magnitudes of the hydroxonium catalytic coefficients remain consis-tent throughout ( Table 2, Experimental section). The rates of hydrolysis for 5 and 7 are essentially the same, indicating that a tert-butyl group at Cα exerts little or no transannular steric demand which might manifest itself in the rate determining step. In the case of Cβ substituted derivatives, a gradual increase in the hydroxonium catalytic coefficient is observed with respect to 5, with a dramatic acceleration noted for the case of 16 (i.e., k H+ 5 = 6.5; 9 ≈ 11; 12 ≈ 14; 16 = 75.9 M −1 s −1 ; Figure 1). Inspection of Newman projections (Figure 4) of 9 and 12 reveal incremental 1,3-transannular steric demand associated with Me-Cβ and C(4/5)-H atoms; consistent with the gradual increase of k H+ . In the case of neopentyl 16 however, two Me-Cβ groups are oriented toward transannular C(4/5)-H atoms at all times, suggesting a means by which this substituent affects a dramatic (>400%) rate increase for this substrate -this is examined further. To understand the significant increase in the hydroxonium catalytic coefficient with varying C(2) substituent, computational conformational analyses of 5 and 16 were performed using Density Functional Theory to optimise the resultant structures (see Experimental section). The half-chair (C 2 ) arrangement was found to be the only stable conformer for the 1,3-dioxolane ring with all other conformers being rotamers about the C(2)-CH 2 R and C(2)-OMe bonds [where R = H (5) or t-Bu (16)]. The orientation of the C(2)O-Me group is ignored from this point as it does not substantially affect the relative energies of the C(2)─CH 2 R rotamers. The potential energy surface for 16 is dominated by the syn arrangement of R with respect to the OMe group (16a and 16b in Figure 5; ΔΔH = 0 and 0.8 kJ/mol, respectively); the rotamer which orients the R group anti with respect to the OMe (16c in Figure 5; ΔΔH = 6 kJ/mol) leads to a pseudo-axial orientation of the OMe group through flattening of the 1,3-dioxolane ring ( Figure 5d); presumably this relieves steric pressure between the t-Bu and C(3/5)-H atoms at a cost of approximately 5.7 kJ/mol higher enthalpy. No such flattened conformer exists for 5.
The calculations are consistent with the model presented earlier in Figure 4. The increasing steric demand of substituent R [i.e., R = H (5); → Et (9); → iPr (12) → t-Bu (16)] leads to transannular clashing with axial C(3)/(4)-H atoms and subsequent flattening of the 1,3-dioxolane ring. For 16 this affords an energetically accessible conformer 16c which resembles the planar geometry anticipated for the transition state (Scheme 1), and should therefore be entropically favoured following the principle of least molecular motion. To confirm whether there was any enthalpic barrier to elimination of the protonated methoxy group, geometric scans for potential transition-states were made using Density Functional Theory calculations, by incrementally increasing and fixing the C(2)─OMe bond length for rotamers 16a-c after protonation, and allowing all other geometry elements to optimise. In all cases, the five-membered ring moved towards the final planar oxonium ion, but no enthalpic barrier was found for the C(2)-OMe bond cleavage. This supports entropic control of this elimination reaction, and it is therefore not surprising that the more planar ring for the anti rotamer 16c would lead to a more rapid elimination of methanol after protonation, consistent with an earlier transition state [29]. Compound 5, which does not have such an accessible flattened ring conformation cannot access this lower entropy trajectory and hence reacts more slowly.

Exclusion of O(1/3)─C(5/4) cleavage
It has been noted that annular strain in trans-2,5-dimethyltetrahydrofurans invokes sufficient charge separation to switch the mechanism of ring cleavage from S N 2 → S N 1 [34]. We therefore sought to exclude the possibility of mechanistic partitioning via some strain-induced O(1/3)-C(5/4) cleavage pathway for the systems under study here. Though the products of hydrolytic attack at C(2) or C(4/5) are constitutionally indistinguishable, the participation of a C(4/5) pathway brought about through intramolecular strain may be detected using H 2  We examined 5 alongside 16 as the former, which hydrolyses at 10% the rate of the latter, is not assumed to experience any significant degree of strain. The 1,3-dioxolan-2-ylium salts derived from 5 (5*, i.e., B R = Me) and 16 ( shift (Δδ C = 0.04 ppm) of the resonance attributed to the carbonyl carbon atom [39,40], consistent with C(2) attack alone. We confirm therefore that the dramatic rate acceleration noted for 16 is not consistent with a change in mechanism [41].

Conclusion
A range of promising FDII candidates based upon acyclic and cyclic geminal ethers, unambiguously organised according to their catalytic coefficients k H+ in CD 3 CN/D 2 O 4:1 has been presented ( Figure 1). As anticipated, orthoacetates possessing ethoxy substituents are the most rapidly hydrolysed acyclic systems. Both entropic and steric effects are believed to account for the relatively fast rates of hydrolysis of cyclic orthoesters with respect to ketals. Rate increases within five-membered endo-cyclic orthoesters accompany increasingly bulky C(2) alkyl substitution. Dramatic increases in the hydroxonium catalytic coefficients for hydrolysis are observed for Cβ branching at C(2), which can lead to conformational distortion of the fivemembered ring which lowers the activation barrier to elimination of the OMe group and formation of the intermediate oxonium ion. The dual performance of cyclic geminal ethers as FDII for jet fuels will be reported shortly.

Experimental
All preparative operations were performed at the synthetic laboratories of the School of Chemistry, University of Bristol. NMR spectra were recorded on a JEOL ECP Eclipse 300 spectrometer at 300 MHz ( 1 H), a JEOL ECP Eclipse 400 spectrometer at 400 MHz ( 1 H) and 100 MHz ( 13 C) or a Varian VNMRS 500 spectrometer at 500 MHz ( 1 H) in deuterated solvents (University of Bristol). Mass spectra (HRMS) were obtained on a Micromass AutoSpec MS in CI mode using CH 4 as the collision gas. HRMS and elemental analysis were performed by the respective services at the School of Chemistry, University of Bristol.
Computational techniques. X-ray crystal structures were located in the 2014 release of the Cambridge Structural Database (CSD v 5.35, which contains 658, 007 entries) using the Conquest software (v 1.16) and visualised using the Mercury software package (v 3.1). Conformational searching was conducted with Macromodel software, using a MonteCarlo search method and the MMFF force field. Final molecular geometries were optimised using the Gaussian09 [49] with density functional level of theory, using the hybrid functional B3LYP/6-31G* to optimise structures [50][51][52], with convergence criteria for maximum and RMS force (0.000450 and 0.000300 in atomic units per Bohr and per radian respectively) and for atomic displacements (0.001800 and 0.001200 Angstroms, respectively).

Supporting Information File 1
Analytical data.